Although many semiconductor devices are called xe2x80x9cswitches,xe2x80x9d and although those devices are used in many circuit applications to perform the electrical connection functions of a traditional metal-to-metal moving contact structure, it is still the case that for a variety of reasons (e.g., ability to carry high currents, high break-down voltages, high isolation, operation in an AC circuit, etc.) a genuine traditional switch is the component of choice. Of course, the term xe2x80x9cswitchxe2x80x9d is broader than the simple class of devices that are operated by a human hand or finger, or by some mechanical linkage to an object such as a door, cockpit canopy or a float, and the term includes what are ordinarily called xe2x80x9crelays.xe2x80x9d A relay is a switch that is (usually) operated by an electrical signal that is converted (e.g., by a magnetic coil) to mechanical motion that operates the switch. Common relays incorporate a spring tension to return the contacts to an un-operated state in the absence of the electrical signal. On the other hand, some relays have actuation mechanisms that transition from one stable state to another stable state, and that stay transitioned in the absence of, or after the removal of, the signal that produced the change in state. Such relays are called xe2x80x9clatchingxe2x80x9d relays.
Among the reasons for preferring a genuine moving contact switch to one of its semiconductor counterparts is the need for preserving the characteristic impedance of a transmission line that must be switched among other components (attenuators, power splitters, etc.), or the need for simple shielding in a less demanding situation of a conductor that is only shielded, but that is not an actual transmission line having a controlled characteristic impedance. xe2x80x9cCoaxial switchxe2x80x9d is the term usually given to this sort of structure, and various instances of this sort of thing are produced as relays also, in both latching and non-latching versions. True coaxial relays are an exercise in electro-mechanical fidelity to the transmission line that they are to connect to. They are not small, and they are not inexpensive. They wear out, their contacts oxidize or deform, and their behavior can become erratic. But most of all, they are xe2x80x9cbigxe2x80x9d and are unsuitable for use in many applications involving integrated circuitry, including the assembly of integrated system components onto a substrate to form a xe2x80x9chybridxe2x80x9d circuit. We simply can""t bring ourselves to use a relay whose volume is ten to even a thousand times that of the circuitry it is supposed to switch, let alone use several such relays!
On the other hand, if a genuine metal-to-metal switching mechanism is small enough, then below some upper frequency it can largely avoid the evils of a temporary (think: xe2x80x9csmallxe2x80x9d) discontinuity or lapse in shielding, even though it is not itself a coaxial structure. This is follows from the well appreciated fact that when the physical size of the departure from ideal geometry is small compared to the shortest wavelength present, then the resulting discontinuity is essentially invisible, or at least tolerable. This is a long way of saying that if we have a genuine switch that is small enough, we may well be able to use it in place of a much larger coaxial switch, even though the small switch is not truly a segment of a transmission line structure. The same goes for shielding, although that is often easier to supply, since it does not have the worry of maintaining a characteristic impedance that is strongly influenced by geometry. And equally as valuable, such a small relay would be of a size that is comparable to, or maybe even a little smaller than, the circuit elements it is to switch among, and the whole works can be fabricated on a substrate as a hybrid. We have just called such a thing a relay, rather than a switch, since at the sizes we are interested in (say, one tenth inch by one tenth inch) it is most unlikely that such a switch would have a bat handle, lever or other mechanical linkage through which it is to be operated. In times past such a relay was fanciful, but that is no longer so.
Recent developments have occurred in the field of very small switches having liquid moving metal-to-metal contacts. With reference to FIGS. 1-4, we shall briefly sketch the general idea behind one class of these devices. Having done that, we shall advance to the topic that is most of interest to us, which is a technique for fabricating on a hybrid substrate dense collections of such relays. (Henceforth we shall, as is becoming customary, refer to such a switch as a Liquid Metal Micro Switch, or LIMMS.)
Refer now to FIG. 1A, which is a top sectional view of certain elements to be arranged within a cover block 2 of suitable material, such as glass. The cover block 2 has within it a closed-ended channel 1 in which there are two small movable distended droplets (10, 11) of a conductive liquid metal, such as mercury. The channel 1 is relatively small, and appears to the droplets of mercury to be a capillary, so that surface tension plays a large part in determining the behavior of the mercury. One of the droplets is long, and shorts across two adjacent electrical contacts extending into the channel, while the other droplet is short, touching only one electrical contact. There are also two cavities 6 and 7, within which are respective heaters 4 and 5, each of which is surrounded by a respective captive atmosphere (15, 16) of an inert gas, such as CO2. Cavity 4 is coupled to the channel 1 by a small passage 8, opening into the channel 1 at a location about one third or one fourth the length of the channel from its end. A similar passage 9 likewise connects cavity 5 to the opposite end of the channel. The idea is that a temperature rise from one of the heaters causes the gas surrounding that heater to expand, which splits and moves a portion of one of the mercury droplets, forcing it to join the short droplet. This forms a complementary physical configuration (or mirror image), with the large droplet now at the other end of the channel. This, in turn, toggles which two of the three electrical contacts are shorted together. After the change the heater is allowed to cool, but surface tension keeps the mercury droplets in their new places until the other heater heats up and drives a portion of the new long droplet back the other way. Since all this is quite small, it can all happen rather quickly; say, on the order of milliseconds.
To continue, then, refer now to FIG. 1B, which is a sectional side view of FIG. 1A, taken through the middle of the heaters 4 and 5. New elements in this view are the bottom substrate 3, which may be of a suitable ceramic material, such as that commonly used in the manufacturing of hybrid circuits having thin film, thick film or silicon die components. A layer 17 of sealing adhesive bonds the cover block 2 to the substrate 3, which also makes the cavities 4 and 5, passages 8 and 9, and the channel 1, all gas tight (and also mercury proof, as well!). Layer 17 may be of a material called CYTOP (a registered trademark of Ashai Glass Co., and available from Bellex International Corp., of Wilmington, Delaware).
Also newly visible are vias 18-21 which, besides being gas tight, pass through the substrate 3 to afford electrical connections to the ends of the heaters 4 and 5. So, by applying a voltage between vias 18 and 19, heater 4 can be made to become very hot very quickly. That in turn, causes the region of gas 15 to expand through passage 8 and begin to force long mercury droplet 10 to separate, as is shown in FIG. 2. At this time, and also before heater 4 began to heat, long mercury droplet 10 physically bridges and electrically connects contact vias 12 and 13, after the fashion shown in FIG. 1C. Contact via 14 is at this time in physical and electrical contact with the small mercury droplet 11, but because of the gap between droplets 10 and 11, is not electrically connected to via 13.
Refer now to FIG. 3A, and observe that the separation into two parts of what used to be long mercury droplet 10 has been accomplished by the heated gas 15, and that the right-hand portion (and major part of) the separated mercury has joined what used to be smaller droplet 11. Now droplet 11 is the larger droplet, and droplet 10 is the smaller. Referring to FIG. 3B, note that it is now contact vias 13 and 14 that are physically bridged by the mercury, and thus electrically connected to each other, while contact via 12 is now electrically isolated.
The LIMMS technique described above has a number of interesting characteristics, some of which we shall mention in passing. They make good latching relays, since surface tension holds the mercury droplets in place. They operate in all attitudes, and are reasonably resistant to shock. Their power consumption is modest, and they are small. They have decent isolation, are reasonably fast with minimal contact bounce. There are versions where a piezo-electrical element accomplishes the volume change, rather than a heated and expanding gas. There are also certain refinements that are sometime thought useful, such as bulges or constrictions in the channel or the passages. Those interested in such refinements are referred to the Patent literature, as there is ongoing work in those areas. See, for example, the incorporated U.S. Pat. No. 6,323,447 B1.
To sum up our brief survey of the starting point in LIMMS technology that is presently of interest to us, refer now to FIG. 4. There is shown an exploded view of a slightly different arrangement of the parts, although the operation is just as described in connection with FIGS. 1-3. In particular note that in this arrangement the heaters (4, 5) and their cavities (6, 7) are each on opposite sides of the channel 1.
Now consider the underside of the substrate 3. For a SPDT switch it will have a pattern of seven vias thereon, or perhaps six if there is a trace on the substrate connecting two of the heater terminals in common. (A SPST device might have a few as five, or as many as six.) It is through these vias (not shown, but it is clear that they are there and where they are) that connection is made to the LIMMS of FIG. 4. Now suppose that we desire to deploy many LIMMS in a small space, or perhaps only few LIMMS, but as closely together as possible, say, owing to routing considerations for microwave signals. There are also shielding and transmission line considerations, all of which suggest that a ground plane is desirable. In addition, we""d like to be able to treat such a collection of LIMMS devices as a unit assembly. What to do?
A solution to the problem of locating a plurality of Liquid Metal Micro Switches (LIMMS) on a substrate and in a minimal amount of space is to mount them on opposite sides of a multi-layer substrate. Vias on the substrate and located within the footprints ofthe LIMMS serve to make connection with the LIMMS. Traces on the internal layers of the multi-layer substrate are routed around and over each other to arrive at a perimeter surrounding the LIMMS, where they emerge again as vias and are available for interconnection with further circuitry via conventional techniques, such as solder balls, wire bonding, a socket, etc. The multi-layer substrate may also incorporate a ground plane to assist in shielding and the fabrication of any interconnecting transmission lines.